Characterization of the Amino-terminal Activation Domain of Peroxisome Proliferator-activated Receptor a IMPORTANCE OF a -HELICAL STRUCTURE IN THE TRANSACTIVATING FUNCTION*

The transactivating function of the A/B region of mouse peroxisome proliferator-activated receptor a (PPAR a ; NR1C1) was characterized. The truncated version of PPAR a lacking the A/B region had 60–70% lower transactivating function than full-length PPAR a in both the presence and absence of the peroxisome proliferator ciprofibrate. When tethered to the yeast Gal4 DNA-bind-ing domain, the A/B region exhibited the significant li-gand-independent transactivating function, AF-1 activity. The first 44 amino acid residues were necessary for maximal transactivation, and the minimally essential region was further delimited to amino acids 15–44. This region is highly enriched with acidic residues, but mu-tational analyses showed that the protein structure, rather than the negative charge itself, was important for the AF-1 activity. An a -helical configuration was predicted for this region, and a CD spectrum analysis of the synthetic peptides showed that mutant sequences with higher AF-1 activity have higher helical contents and vice versa. The most active mutant, in which Met 31 was replaced with Leu, was ; 5-fold more potent than the wild-type A/B region. These findings indicate that the AF-1 region of PPAR a is an acidic activation domain and that the helix-forming property is implicated in the transactivating function. Jasco J-720 spectrometer pH 6.8 presence of 0–40% trifluoroethanol (TFE).

Peroxisome proliferator-activated receptors (PPARs) 1 constitute a subfamily (NR1C, according to the unified nomenclature (1)) of nuclear hormone receptors. Since the first cloning of PPAR␣ (NR1C1) from mouse (2), three isoforms (␣, ␤ (or ␦), and ␥) have been identified in many organisms, including human, rat, and Xenopus (for review, see Ref. 3). Many studies suggest that PPAR␣ regulates the fatty acid metabolism by controlling the expression of the genes involved in fatty acid oxidation as well as lipoprotein subunits, positively or negatively (3). On the other hand, PPAR␥ (NR1C3) seems to govern more versatile physiological processes such as adipocyte differentiation (4), inflammatory response (5,6), and maturation of macrophages into foam cells (7,8). Very recently, evidence showing the involvement of PPAR␦ (NR1C2) in embryo implantation was presented (9).
Many compounds have been identified as ligands of PPAR (3). The fibrate class of peroxisome proliferators and leukotriene B 4 are relatively selective for PPAR␣, whereas the thiazolidinedione class of antidiabetic compounds, 15-deoxy-⌬ 12,14prostaglandin J 2 , and anti-inflammatory agents are specific for PPAR␥. Unsaturated long-chain fatty acids are probably important endogenous ligands common to PPAR␣ and PPAR␥.
As in most nuclear receptors, PPAR comprises four discrete functional domains, A/B, C, D, and E, in the order from the amino to the carboxyl terminus (15). In general, the C region is highly conserved among the nuclear receptor superfamily and contains two zinc finger motifs involved in DNA binding. The E region is the next most highly conserved and contains the ligand-binding site and ligand-dependent activation domain (AF-2). The D region is a hinge domain between the C and E regions and contains sequences important for heterodimerization and the AF-2 activity. The A/B region is the most variable among nuclear receptors and, in many cases, has a ligandindependent gene-activating function (AF-1). The AF-1 region acts independently of AF-2 when tethered to heterologous DNA-binding domains and sometimes synergizes with AF-2 through intramolecular interaction. Regulation of the receptor activities by growth factors and other extracellular signals is often mediated through mitogen-activated protein kinase-dependent phosphorylation of specific sites in the AF-1 region.
For PPAR, AF-1 activities were found in the A/B region of the ␥-isoform (16,17). Mitogen-activated protein kinase-dependent phosphorylation in the A/B region has also been demonstrated and is implicated in the actions of insulin and growth factors on adipocyte differentiation and other PPAR␥ functions (16, 18 -23). The AF-1 activation domain, however, has not yet been mapped precisely in the A/B region. In this study, we identified and characterized the AF-1 domain in the A/B region of PPAR␣. This domain is enriched with acidic amino acids. Mu-tational analysis combined with CD spectrum analysis of synthetic peptides showed that the helix-forming property, rather than the negative charge itself, is important for the activating function.

EXPERIMENTAL PROCEDURES
Plasmids-The expression vector of full-length PPAR␣ was pNCM-VPPAR␣ (11), constructed by replacing the lacZ gene in the cytomegalovirus promoter-driven vector pCMV␤ (24) with mouse PPAR␣ cDNA. A deletion construct lacking the A/B region was created by inserting a double-strand linker composed of 5Ј-TGTCGAATATGTGGGGA-CAAGG-3Ј and 5Ј-CCTTGTCCCCACATATTCGACACATG-3Ј between the BstXI site encompassing the initiation codon and the StuI site in the early portion of the C domain-encoding region. The resulting plasmid, pNCMVPPAR␣⌬A/B, codes for a protein sequence in which the initiation methionine is directly linked to the first cysteine residue (Cys 102 ) of the C domain.
For constructing the fusion of yeast Gal4-BD and the A/B domain of PPAR␣, the cDNA sequence of the corresponding region was amplified by polymerase chain reaction using an upstream primer (5Ј-CCAGAAT-TCATGGTGGACACAGAGAGCCCC-3Ј) and a downstream primer (5Ј-CGTTGATCACTCGATGTTCAGGGCACTGCC-3Ј). The amplified fragment was digested with EcoRI and BclI, yielding two fragments of 76 and 227 base pairs due to the presence of an internal EcoRI site. The longer fragment was first inserted between the EcoRI and BamHI sites of the plasmid pCMX-Gal4-N, which encodes a yeast Gal4-BD (amino acids 1-147) and is driven by the cytomegalovirus promoter. The shorter fragment was then inserted at the EcoRI site, and a clone containing the fragment in the correct orientation was verified by nucleotide sequencing.
The A/B region-coding sequences of mouse PPAR␥2 and PPAR␦ were also amplified by polymerase chain reaction and inserted in pCMX-Gal4-N. The cDNA sequence of the PPAR␥1 A/B region was created from the PPAR␥2 sequence, taking advantage of PPAR␥2, which has 30 extra amino acid residues at the N terminus compared with the PPAR␥1 sequence, and also, the sequence C-terminal to the second methionine (Met 31 ) is identical to that of PPAR␥1. Thus, the region from the NcoI site encompassing the Met 31 codon to the position corresponding to the end of the PPAR␥2 A/B domain was cleaved and inserted in pCMX-Gal4-N.
For assaying the transcriptional activation by PPAR constructs containing the C region, a reporter vector, pAOXPPREluc (11), containing the PPAR-binding site and the basal promoter derived from the rat acyl-CoA oxidase gene was used (25). For assaying the activation by Gal4 fusion constructs, tk-GALpx3-luc, containing three copies of the Gal4-binding sites (upstream activation site (UAS)) and the herpes simplex virus thymidine kinase promoter, was mostly used. tk-GALpx1-luc, containing one copy of the Gal4-binding site, was also used in some experiments. This construct was accidentally produced by recombination during amplification in bacteria.
Construction of Mutants-Deletion constructs of the PPAR␣ A/B region were produced by cutting pCMX-Gal4-PPAR␣A/B with restriction enzymes having recognition sites in the A/B region-coding sequence. These enzymes were EcoRI (encompassing the codons for Glu 26 and Phe 27 ), SacI (Ser 45 ), ScaI (Tyr 56 ), and PvuII (Ser 82 ). DNA sequence coding for the portion C-terminal to amino acid residue 15 was created by polymerase chain reaction amplification.
Site-directed mutagenesis was carried out using the QuickChange mutagenesis kit (Stratagene) according to the manufacturer's protocol. All mutations were confirmed by nucleotide sequencing.
DNA Transfection-The HeLa human cervical cancer cell line was used in most experiments, whereas the HepG2 human hepatoma line and the CV-1 monkey kidney cell line were used in some cases. Cells were cultured in Dulbecco's modified Eagle's medium or Ham's F-12 medium containing 10% fetal bovine serum.
Transfection was carried out by the calcium phosphate method (26) with slight modifications. For transfection, 1 ϫ 10 5 cells were seeded in 60-mm dishes and cultured overnight. The following day, each dish received DNA/calcium phosphate precipitates containing 1 g of the expression vector of a PPAR construct, 4 g of an appropriate luciferase reporter plasmid, and 4 g of a ␤-galactosidase expression vector (pCMV␤) as a reference. After 4 h, the precipitates were removed, and the cells were cultured for 40 h in the presence or absence of 0.5 mM ciprofibrate, a peroxisome proliferator. Cells were lysed, and luciferase and ␤-galactosidase activities were measured as described previously (11). Luciferase activity was normalized for transfection efficiency based on the ␤-galactosidase activity. Three or more independent as-says were carried out for most series of experiments, and the results are expressed as mean relative values Ϯ S.D.
Synthesis of Peptides and CD Spectrum Measurement-To analyze the secondary structure of the N-terminal region of PPAR␣, a peptide corresponding to amino acids 21-39 (SPLSEEFLQEMGNIQEISQ) was synthesized on an automated solid-phase peptide synthesizer (Shimadzu PSSM-8) using Tenta Gel TG-RAM resin and Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry. Three extra amino acids (GGYamide) were attached to the C terminus to make the spectrophotometric determination of peptide concentration feasible without affecting the CD signal (27). Mutant peptides containing amino acid substitutions were also synthesized. Peptides were purified by reverse-phase high performance liquid chromatography on a C 18 column, and their respective molecular masses were confirmed by mass spectrometry on a timeof-flight mass spectrometer (Shimadzu/Kratos Kompact MALDI II). The CD spectra of these peptides were recorded on a Jasco J-720 CD spectrometer at pH 6.8 in the presence of 0 -40% trifluoroethanol (TFE).

Effect of Deletion of the A/B Region from Full-length PPAR␣-
We first examined whether the transcriptional activating function is affected when the A/B region is deleted from PPAR␣. When the full-length PPAR␣ expression vector was transfected together with the luciferase reporter vector carrying a single copy of the peroxisome proliferator response element of rat acyl-CoA oxidase, the luciferase expression was enhanced 15-fold as compared with the value obtained with an empty expression vector, pCMVNot ( Fig. 1, bars 1 and 5). The luciferase expression was further stimulated nearly 4-fold in the presence of ciprofibrate, a peroxisome proliferator (Fig. 1,  bar 6). When PPAR␣ with the A/B region deleted was expressed, the luciferase activity decreased by 60 -70% in either the presence or absence of ciprofibrate ( Fig. 1, bars 3 and 4). Thus, deletion of the A/B region from PPAR␣ decreased the transcriptional activating function ϳ3-fold without affecting inducibility by ciprofibrate.
AF-1 Activity of the A/B Region-We next investigated whether the PPAR␣ A/B region has a ligand-independent activating function (AF-1). For this purpose, we constructed plasmids in which the coding sequence of the A/B region as well as full-length PPAR␣ was tethered to the Gal4-BD-coding sequence. When the fusion construct of full-length PPAR␣ was cotransfected with a luciferase reporter vector containing three copies of Gal4-UAS (tk-GALpx3-luc), strong transcriptional activation dependent on ciprofibrate was observed ( Fig. 2A). The scale of ligand dependence was much larger than that obtained with unfused PPAR␣ on the reporter gene driven by the peroxisome proliferator-response element (compare Figs. 1 and 2A). On the other hand, the fusion construct of the A/B region and Gal4-BD exhibited ϳ10-fold activation as compared with the activity of simple Gal4-BD in either the presence or absence of ciprofibrate. Similar results were obtained with tk-GALpx1luc, which contains a single copy of UAS (data not shown).
We compared the AF-1 activities of the PPAR␣, PPAR␥1, PPAR␥2, and PPAR␦ isoforms, all of mouse origin (Fig. 2B). Under the assay conditions, the AF-1 activity of PPAR␣ was significantly higher than that of other isoforms. The A/B region of PPAR␥2 exhibited a small AF-1 activity, but the activity were negligible for PPAR␥1 and PPAR␦. Much higher AF-1 activity was reported for PPAR␥2 compared with PPAR␥1 (16).
Localization of the AF-1 Activity in the N-terminal 44 Residues-We tried to narrow down the PPAR␣ AF-1 region in the Gal4-BD fusion context (Fig. 3). When the A/B region sequence was deleted stepwise from the C-terminal side starting at amino acid 101, the construct containing amino acids 1-44 exhibited almost full AF-1 activity, whereas that containing amino acids 1-27 did not. On the other hand, deletion of 14 residues from the N-terminal side decreased the reporter expression by 70%, and deletion of the first 25 residues completely removed the activity. Hence, AF-1 activity is assigned to the N-terminal 44-amino acid sequence, in particular to resi-dues 15-44.
We confirmed for representative constructs that the fusion proteins were correctly expressed in the transfected cells. Proteins of the expected sizes were detected by Western blotting with anti-Gal4-BD antibody for Gal4-BD and Gal4 fusion constructs with the PPAR␣ A/B region (amino acids 1-27 and 1-55) (Fig. 4, lanes 1-4). The expression levels were comparable except for the wild-type PPAR␣ A/B region fusion construct, which exhibited significantly lower expression reproducibly in repeated experiments.
PPAR␣ AF-1 Is an Acidic Activation Domain-The amino acid sequence of the A/B region of PPAR␣ is shown in Fig. 5A. This region consists of 101 amino acid residues and contains 18 acidic residues, but no basic residues. The region of amino acids 15-44 that was minimally essential for transactivation was particularly enriched with acidic residues in which 10 out of 30 residues were acidic. Thus, the AF-1 region of PPAR␣ seemed to be a typical acidic activation domain.
To examine the roles of acidic residues, we performed point mutation studies on the Gal4-BD-PPAR␣A/B background. We changed doublets of acidic residues (Asp 17 -Asp 18 , Glu 25 -Glu 26 , and Glu 43 -Glu 44 ) to asparagine doublets (mutants dm1-3) (Fig.  5A) and also made combinations of these mutations. Thus, up to six negative charges were removed from the original 10. We did not use glutamine for glutamic acid to avoid the possible creation of a glutamine-rich activation domain. Series of asparagine substitutions were successfully employed for a viral acidic activator, VP16 (28).
Individual doublet mutations of the acidic residues had only a marginal effect on AF-1 activity (Fig. 6A). Combinations of two doublet mutations exhibited a significant decrease in AF-1 activity compared with the AF-1 activity of the wild-type A/B region, particularly the combination of dm1 and dm2. The effect of dm3 seemed smaller than those of dm1 and dm2. The combination of the three doublet mutations resulted in only a slightly larger decrease in the activity compared with the dm1/ dm2 combination, and even this mutant still retained a significant level of transcriptional activation as compared with Gal4-BD (2.6-fold). We confirmed these results with the (UAS) 1 reporter and also with two other cell lines, HepG2 and CV-1 (data not shown). The results support the importance of acidic residues for AF-1 activity, but the effects of mutations were unexpectedly small if the negative charges themselves are important.
Implication of Protein Structure in the AF-1 Activity-The secondary structure of the PPAR␣ A/B region predicted by the Chou-Fasman method (29) is shown in Fig. 5B. Strikingly, almost the whole region minimally essential for AF-1 activity was deduced to take the ␣-helical configuration. In helical wheel analysis (30) of region acid 15-44, most of the acidic amino acids were found on one side, whereas the hydrophobic residues were on the other (Fig. 5C). Thus, this region seemed to form an amphiphilic ␣-helix. Doublet mutations dm1 and dm2 were likely to break the helical structure locally, and the combination of dm1 and dm2 expanded the unstructured region additively (Fig. 5B). On the other hand, mutation dm3, which was less effective in decreasing the AF-1 activity in the transfection assay, did not seem to affect the configuration of the main part of the AF-1 domain, either by itself or in combination with other mutations. These predictions led us to assume that the protein structure, rather than the negative charges, in this region was important.
To examine this possibility, we searched for other mutations that were expected to break the ␣-helix based on the Chou-Fasman prediction. We found that Met 31 contributes to a great extent to the helix-forming probability. Substitution of Met 31 with Gly or Ala was predicted to change the configuration between residues 31 and 38 from ␣-helix to ␤-sheet (Fig. 5B,  bottom). Accordingly, we made fusion proteins of M31G and M31A mutant versions of the PPAR␣ A/B region and Gal4-BD. The AF-1 activity was mostly lost in both mutants, with the effect of M31G being more severe than that of M31A (18.6 and 26.9% active as compared with the wild type, respectively) (Fig.  6B). We also investigated the effects of combinations of the M31G or M31A mutation with dm1, dm2, or dm3. The dm series of mutations resulted in no further decrease in the AF-1 activity over the M31G or M31A mutation (data not shown). We confirmed that the M31G, M31A, and dm3 fusion proteins were expressed at comparable levels in the cells upon transfection of the respective expression vectors as compared with Gal4-BD and two deletion mutants (Fig. 4).
We next examined whether these mutations indeed broke the ␣-helical structure by CD spectrum analysis of the synthetic peptides corresponding to the wild-type and mutant AF-1 sequences. We first predicted the ␣-helical contents of a series of peptides using the AGADIR algorithm (31) to select the peptide sequences to be synthesized. The peptide sequence 21-39 attached with the GGY tripeptide at the C terminus was expected to have an ␣-helical content of 4.05% at pH 7.0 (Fig.  7A), whereas M31G and M31A exhibited ␣-helical contents of 1.00 and 1.73%, respectively. The relative order in helical contents of these peptides did not change when calculation was performed on longer peptide sequences, and the highest values were obtained with the above segment. The E25N/E26N double mutant (dm2) was also expected to have a significantly lower ␣-helical content. Conversely, we found that the M31L and E26R mutants have the highest predicted helical contents among the possible mutants at residues 31 and 26, respectively. The E26R/M31L double mutant was expected to have an even higher helical content.
Based on these predictions, we synthesized the above peptides and measured the CD spectra in the presence and absence of TFE (Fig. 7B). Because the absolute ellipticity values at 222 nm ([] 222 ) of the peptides in the absence of TFE were too small to be compared, increasing concentrations of TFE, which is known to stabilize native-like helical structures by strengthening peptide hydrogen bonds, were added to evaluate the helixforming properties of these peptides. The difference in [] 222 among peptides was expanded at higher TFE concentrations. The E25N/E26N double mutant (dm2) responded to TFE in an unusual manner, exhibiting almost no spectral change around 222 nm, even when the TFE concentration was raised to 40%. This mutant peptide might have an abnormal secondary structure in solution. The three other mutant peptides were also analyzed, and the calculated ␣-helical contents at 30% TFE are shown in Fig. 7C. As expected, M31G and M31A had lower ␣-helical contents and M31L had a higher ␣-helical content compared with the wild type. On the other hand, mutation E26R did not increase the ␣-helical content, contrary to the prediction using the AGADIR algorithm. Thus, mutant E26R had a slightly lower ␣-helical content compared with the wild type, whereas the value of E26R/M31L was intermediate between those of the wild type and M31L.
Based on these results, we produced mutant Gal4-BD-PPAR␣A/B fusion constructs carrying the mutations M31L, E26R, and E26R/M31L. In the transfection assay, M31L had a strikingly higher AF-1 activity than the wild type (Fig. 6B). In contrast, E26R had a slightly lower AF-1 activity than the wild type, whereas E26R/M31L had an activity intermediate between those of M31L and the wild type. Thus, change in the charge by ϩ2 in mutant E26R did not drastically affect the transcriptional activating function, as in the case of mutant dm2. We plotted the AF-1 activities of wild-type and mutant constructs against the ␣-helical contents of corresponding peptides (Fig. 7C). AF-1 activity correlated well with the helical content following an exponential function, suggesting the importance of the ␣-helix-forming property of this region.
Effects of M31G and M31L Mutations in the Whole PPAR Sequence-We introduced the mutations into the whole PPAR␣ protein sequence. When the expression vectors of mutant PPARs were cotransfected with the reporter vector driven by a peroxisome proliferator response element, M31G exhibited a lower transactivation than wild-type PPAR␣ in both the presence and absence of the peroxisome proliferator (Fig. 1). The M31L mutation yielded a slight increase in activity compared with the wild type, and the difference between the activities of M31L and M31G was quite significant. The effects of these mutations in the intact PPAR␣ context were smaller than those in the Gal4 fusion context, possibly due to the structural constraint. The activities of the wild type and the two mutants, however, were in the same order as those of the Gal4 fusion constructs of the A/B region mutants. DISCUSSION These results indicate that a ligand-independent transactivating function is located in the A/B region of mouse PPAR␣. The sequence important for the activity is restricted to region 1-44, particularly amino acids 15-44. After completion of this study, the AF-1 activity of human PPAR␣ and its regulation through phosphorylation were reported (32), but the minimally essential region was not specified.
The main part of the AF-1 region is highly enriched with acidic residues and devoid of any basic residue. Thus, this region seems to be a typical acidic activation domain. The results of mutation studies, however, raised a question concerning the importance of the negative charge itself. Rather, these results together with the CD spectrum analysis of synthetic peptides suggest the importance of the ␣-helix-forming properties of this region, i.e. the mutations resulting in a higher helical content gave higher AF-1 activity and vice versa. Helical wheel analysis suggests that this region takes an amphiphilic ␣-helix. These characteristics are consistent with reported results on acidic activation domains of other proteins: the importance of the negative charge itself was not supported by mutation studies on the viral activator VP16 (33). Involvement of the ␣-helical structure induced by the interaction with one of the TATA-binding protein associated factors, TAF II 31, was shown by an NMR study on VP16 (34). The importance of the amphiphilic ␣-helix was also shown in a crystallographic study on the complex of the acidic activation domain of p53 and the attenuator protein MDM2 (35). The TFE-induced formation of the ␣-helix in the PPAR␣ AF-1 region possibly mimics the structural change induced by a protein-protein interaction.
Comparison of several acidic activation domains as well as mutation analysis led to a proposal regarding the importance of the FXX motif (34), where denotes a hydrophobic amino acid. In PPAR␣, sequence 27-31 corresponds with this motif at Phe 27 and Met 31 , but not at Glu 30 . The importance of the primary structure including several aromatic and hydrophobic residues was proposed based on sequence comparison among the AF-1 regions of nuclear receptors (36,37). The AF-1 region of PPAR␣ contains only one aromatic residue (Phe 27 ) and has no significant similarity to the proposed consensus sequence. Thus, the AF-1 region of PPAR␣ seems to share the proposed characteristics only imperfectly with other acidic activation domains.
For PPAR isoforms, phosphorylation at the mitogen-activated protein kinase sites in the A/B regions has been highlighted in relation to AF-1 activity. PPAR␣ was shown to be a phosphoprotein (38), and very recently, the phosphorylation sites were mapped at Ser 12 and Ser 21 (32). It has also been shown that phosphorylation is promoted by insulin, potentiating the AF-1 activity. The second phosphorylation site (Ser 21 ) is located within the essential region of AF-1 characterized in this study. How phosphorylation at these sites affects the protein structure of the AF-1 region of PPAR␣ would be an interesting issue. On the other hand, phosphorylation of Ser 82 of PPAR␥1 or Ser 112 of PPAR␥2 by the mitogen-activated protein kinase family has been reported by many researchers (16, 18 -23). Most of them (18 -22) argued that phosphorylation of the PPAR␥ A/B region suppresses the transcriptional activating function of the receptor and leads to decreased differentiation of preadipocytes, a major physiological consequence of PPAR␥ activation. Others (16,22), however, presented contradictory results, claiming potentiation of the activating function of PPAR␥ as well as promotion of adipogenesis by phosphorylation. Ligand-independent AF-1 activities were shown in the A/B regions of PPAR␥1 and PPAR␥2 (16), the latter being markedly more active. On the other hand, Shao et al. (22) documented an interdomain communication between the A/B and ligand-binding domains of PPAR␥, resulting in modulation of ligand-binding affinity.
The AF-1 activities of many nuclear receptors have significant functions in regulating the transcriptional regulatory activity, not simply by conferring the activating function in the absence of ligands, but more important, by synergizing with AF-2. Functional interactions between AF-1 and AF-2, direct or indirect, were observed in several instances (e.g. Ref. 39 -41). Interaction of AF-1 regions with coactivators has also been reported for several nuclear receptors (17,37,42). Comparison of Figs. 1 and 2 suggests that PPAR␣ AF-1 also synergistically activates transcription together with AF-2, as found in other nuclear receptors. The identification of interaction partners and the examination of possible intramolecular communication of PPAR␣ are the subjects of future studies.